Bounty for the first Open Source driver for Kinect

Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

Hi all,
If you still not aware of this:

People offers 2000$ for the first Kinect open source driver working on Linux.
http://goo.gl/jMjsd

Announcement from the original site :
http://goo.gl/3uINX

We may have some people interested in this ML.

Re: Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

MS Kinect interfacing via libusb released
http://www.youtube.com/watch?v=rKhW-cvpkks

http://git.marcansoft.com/?p=libfreenect.git

i

Re: Bounty for the first Open Source driver for Kinect

[Antonio Ospite]

[-- Attachment #1: Type: text/plain, Size: 745 bytes --]

On Wed, 10 Nov 2010 15:20:40 +0100
Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:

> MS Kinect interfacing via libusb released
> http://www.youtube.com/watch?v=rKhW-cvpkks
> 
> http://git.marcansoft.com/?p=libfreenect.git
> 

Good, if anyone is willing to provide the hardware I think I can help
with a proper gspca driver (I helped with the PS3 Eye already). Are
there other RGB-Depth cams supported in linux? Are they usually exposed
just as two distinct cameras?

Regards,
   Antonio

-- 
Antonio Ospite
http://ao2.it

PGP public key ID: 0x4553B001

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: Bounty for the first Open Source driver for Kinect

[Steven Toth]

On 11/10/10 10:54 AM, Antonio Ospite wrote:
> On Wed, 10 Nov 2010 15:20:40 +0100
> Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:
> 
>> MS Kinect interfacing via libusb released
>> http://www.youtube.com/watch?v=rKhW-cvpkks
>>
>> http://git.marcansoft.com/?p=libfreenect.git
>>
> 
> Good, if anyone is willing to provide the hardware I think I can help
> with a proper gspca driver (I helped with the PS3 Eye already). Are
> there other RGB-Depth cams supported in linux? Are they usually exposed
> just as two distinct cameras?

Antonio,

Excellent!

If you are willing to donate your personal time freely to Linux then Kernel Labs
are willing to assist by shipping a Kinect unit to you.

Let me be clear, Kernel Labs have no commercial interest in this project. We
simply like to encourage Linux media projects where possible. Projects like this
are good for Linux and thus good for the community.

My only requirement is that you post regular project status emails to this
mailing list so we can all benefit from your thoughts, rants, any problems or
comments on the project! :)

Drop me a private email if you're interested.

Regards,

- Steve

-- 
Steven Toth - Kernel Labs
http://www.kernellabs.com

Re: Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

Thanks for your interest.

The developers are connected to #openkinect channel on Freenode.
The one who shipped this code is marcan on that channel.


i

Re: Bounty for the first Open Source driver for Kinect

[Antonio Ospite]

[-- Attachment #1: Type: text/plain, Size: 1835 bytes --]

On Wed, 10 Nov 2010 11:11:59 -0500
Steven Toth <stoth@kernellabs.com> wrote:

> On 11/10/10 10:54 AM, Antonio Ospite wrote:
> > On Wed, 10 Nov 2010 15:20:40 +0100
> > Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:
[...]
> >> http://git.marcansoft.com/?p=libfreenect.git
> >>
> > 
> > Good, if anyone is willing to provide the hardware I think I can help
> > with a proper gspca driver (I helped with the PS3 Eye already).
[...]

> Excellent!
> 
> If you are willing to donate your personal time freely to Linux then Kernel Labs
> are willing to assist by shipping a Kinect unit to you.
>
> Let me be clear, Kernel Labs have no commercial interest in this project. We
> simply like to encourage Linux media projects where possible. Projects like this
> are good for Linux and thus good for the community.
>

We share the same view then.

> My only requirement is that you post regular project status emails to this
> mailing list so we can all benefit from your thoughts, rants, any problems or
> comments on the project! :)
>

That sounds fair, emails will be. I stepped forward because even if I
don't want to invest too much time in the protocol analysis phase I can
do the integration work, so that we can get a gspca_kinect driver
eventually; if "eventually" is a good enough deadline for you as a
sponsor then we are set. Obviously with some collaboration the driver
will be delivered sooner than "eventually", and I am pretty sure that is
going to come ;)

> Drop me a private email if you're interested.
>

On its way.

Regards,
   Antonio

-- 
Antonio Ospite
http://ao2.it

PGP public key ID: 0x4553B001

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: Bounty for the first Open Source driver for Kinect

[Andy Walls]

IFixit has a tear down mentioning all the chips and a reference design

http://www.ifixit.com/Teardown/Microsoft-Kinect-Teardown/4066/1

There's two camera sensors, an ir illuminator, four microphones, a motor, and apparently a built in usb hub.

It supposedly consumes more power than a standard usb port can supply, so you may need an external power supply.

Have fun.

Andy

Antonio Ospite <ospite@studenti.unina.it> wrote:

>On Wed, 10 Nov 2010 15:20:40 +0100
>Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:
>
>> MS Kinect interfacing via libusb released
>> http://www.youtube.com/watch?v=rKhW-cvpkks
>> 
>> http://git.marcansoft.com/?p=libfreenect.git
>> 
>
>Good, if anyone is willing to provide the hardware I think I can help
>with a proper gspca driver (I helped with the PS3 Eye already). Are
>there other RGB-Depth cams supported in linux? Are they usually exposed
>just as two distinct cameras?
>
>Regards,
>   Antonio
>
>-- 
>Antonio Ospite
>http://ao2.it
>
>PGP public key ID: 0x4553B001
>
>A: Because it messes up the order in which people normally read text.
>   See http://en.wikipedia.org/wiki/Posting_style
>Q: Why is top-posting such a bad thing?

Re: Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

The bounty is already taken by that developer.

But now, the Kinect thing is supported like a GPL userspace library.
Maybe still need more work to be rewritten as a kernel module.

The device has also a microphone (still need to be hacked), an
accelerometer and even a motor/engine.
The design should be similar to the ps3eye (but 2 video output).
The engine controller and the accelerometer needs to be adressed to
which place ? Linux-input ?.


i

Re: Bounty for the first Open Source driver for Kinect

[Antonio Ospite]

[-- Attachment #1: Type: text/plain, Size: 1398 bytes --]

On Wed, 10 Nov 2010 21:54:58 +0100
Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:

> The bounty is already taken by that developer.
>

Which surely deserves it :)

> But now, the Kinect thing is supported like a GPL userspace library.
> Maybe still need more work to be rewritten as a kernel module.
>
> The device has also a microphone (still need to be hacked), an
> accelerometer and even a motor/engine.
> The design should be similar to the ps3eye (but 2 video output).

PS3 Eye uses bulk transfers while I see Kinect uses iso transfers,
anyway gspca handles both so in the end, yes, the driver could be quite
similar, and the packet scanning routine could be taken from the
userspace library if the license allows that.

> The engine controller and the accelerometer needs to be adressed to
> which place ? Linux-input ?.
>

I think so, exposing the accelerometer as an event device sounds
natural, about the motor I still don't know. Do those show up as usb
HID devices somehow?

About integrating the audio part (which has not even been guessed yet)
is where I shamelessly show my ignorance :)

Regards,
   Antonio

-- 
Antonio Ospite
http://ao2.it

PGP public key ID: 0x4553B001

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: Bounty for the first Open Source driver for Kinect

[Fernando Cassia]

On Wed, Nov 10, 2010 at 6:11 PM, Antonio Ospite
<ospite@studenti.unina.it> wrote:
> On Wed, 10 Nov 2010 21:54:58 +0100
> Mohamed Ikbel Boulabiar <boulabiar@gmail.com> wrote:
>
>> The bounty is already taken by that developer.
>>
>
> Which surely deserves it :)

And then there´s the clueless Ziff-Gates writers who call this Piracy...
http://www.zdnet.com/blog/open-source/is-adafruit-bounty-open-source-or-piracy/7728

FC

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Wed, Nov 10, 2010 at 9:54 PM, Mohamed Ikbel Boulabiar
<boulabiar@gmail.com> wrote:
> The bounty is already taken by that developer.
>
> But now, the Kinect thing is supported like a GPL userspace library.
> Maybe still need more work to be rewritten as a kernel module.
>

This should better remain in userspace and interface libv4l/libv4l2 no
need to make things more complicated than they have to be.

Markus

Re: Bounty for the first Open Source driver for Kinect

[Antonio Ospite]

[-- Attachment #1: Type: text/plain, Size: 1082 bytes --]

On Wed, 10 Nov 2010 22:14:36 +0100
Markus Rechberger <mrechberger@gmail.com> wrote:

> On Wed, Nov 10, 2010 at 9:54 PM, Mohamed Ikbel Boulabiar
> <boulabiar@gmail.com> wrote:
> > The bounty is already taken by that developer.
> >
> > But now, the Kinect thing is supported like a GPL userspace library.
> > Maybe still need more work to be rewritten as a kernel module.
> >
> 
> This should better remain in userspace and interface libv4l/libv4l2 no
> need to make things more complicated than they have to be.

I can see at least two reasons for a kernel driver:
 1. performance
 2. out-of-the-box experience: the casual user who wants to just use
    kinect as a normal webcam doesn't have to care about installing some
    library

If there are arguments against a kernel driver I can't see them yet.

Ciao,
   Antonio

-- 
Antonio Ospite
http://ao2.it

PGP public key ID: 0x4553B001

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
<ospite@studenti.unina.it> wrote:
> On Wed, 10 Nov 2010 22:14:36 +0100
> Markus Rechberger <mrechberger@gmail.com> wrote:
>
>> On Wed, Nov 10, 2010 at 9:54 PM, Mohamed Ikbel Boulabiar
>> <boulabiar@gmail.com> wrote:
>> > The bounty is already taken by that developer.
>> >
>> > But now, the Kinect thing is supported like a GPL userspace library.
>> > Maybe still need more work to be rewritten as a kernel module.
>> >
>>
>> This should better remain in userspace and interface libv4l/libv4l2 no
>> need to make things more complicated than they have to be.
>
> I can see at least two reasons for a kernel driver:
>  1. performance
>  2. out-of-the-box experience: the casual user who wants to just use
>    kinect as a normal webcam doesn't have to care about installing some
>    library

out of the box experience libusb works everywhere, ARM/MIPS/PPC/etc.
Kerneldrivers are usually not installed with those systems.
Higher backward compatibility as well (shall go down to 2.6.15) with
one compiled driver, relevant endusers do not want to compile believe
me. Developers might want but that's another story.

Markus

>
> If there are arguments against a kernel driver I can't see them yet.
>
> Ciao,
>   Antonio
>
> --
> Antonio Ospite
> http://ao2.it
>
> PGP public key ID: 0x4553B001
>
> A: Because it messes up the order in which people normally read text.
>   See http://en.wikipedia.org/wiki/Posting_style
> Q: Why is top-posting such a bad thing?
>

Re: Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
<ospite@studenti.unina.it> wrote:
> If there are arguments against a kernel driver I can't see them yet.

+1
This device is a webcam+(other things), it should be handled similar
to other webcams already supported inside the kernel.
If we make an exception now, we should make many special hacks for
this only case to support it through the other libraries and layers of
the system.

If I want to use this device, I will add many userspace code to create
the skeleton model and that need much computation. Kernel Module adds
performance to my other code.

i

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Wed, Nov 10, 2010 at 11:48 PM, Mohamed Ikbel Boulabiar
<boulabiar@gmail.com> wrote:
> On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
> <ospite@studenti.unina.it> wrote:
>> If there are arguments against a kernel driver I can't see them yet.
>
> +1
> This device is a webcam+(other things), it should be handled similar
> to other webcams already supported inside the kernel.
> If we make an exception now, we should make many special hacks for
> this only case to support it through the other libraries and layers of
> the system.
>
> If I want to use this device, I will add many userspace code to create
> the skeleton model and that need much computation. Kernel Module adds
> performance to my other code.

just some experience from our side, we do have fully working
video4linux1/2 drivers
in userspace, the only exception we have is a very thin layered
kernelmodule in order
to improve the datatransfer. BUT on the other side someone also has to think
about the bandwidth - this is the main consideration why we only use
this for low bandwidth
DVB. Is it worth to optimize any datatransfer that already requires >150 mbit
for embedded devices? (most of them cannot handle analog TV/high
bandwidth cameras due
that high bandwidth requirement).
There's definitely a point speed VS userfriendlyness.
Due high bandwidth need it's kinda obsolete to focus on high
optimizations for this, thus it's better
to focus on higher compatibility (this shall work with what you have
done now already).

I've seen alot projects failing due not having enough users
If it should mainly remain a hacker only project then a kernel module
should be fine.

Aside of that libusb has been ported to BSD and OSX as well, so it
widens the possibilities to other
operating systems. Also think about you update your work and it can
immediately be tested with all
systems without too much work, getting the right kernel sources,
kernel module etc.

But of course it's up to the developer :-) I just want to give some
ideas, since we already went through
with some projects that and I can tell it works.

BR,
Markus

>
> i
>

Re: Bounty for the first Open Source driver for Kinect

[Antonio Ospite]

[-- Attachment #1: Type: text/plain, Size: 1092 bytes --]

On Thu, 11 Nov 2010 00:13:09 +0100
Markus Rechberger <mrechberger@gmail.com> wrote:

> On Wed, Nov 10, 2010 at 11:48 PM, Mohamed Ikbel Boulabiar
> <boulabiar@gmail.com> wrote:
> > On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
> > <ospite@studenti.unina.it> wrote:
> >> If there are arguments against a kernel driver I can't see them yet.
[...]
> > If I want to use this device, I will add many userspace code to create
> > the skeleton model and that need much computation. Kernel Module adds
> > performance to my other code.
> 
> just some experience from our side, we do have fully working
> video4linux1/2 drivers
> in userspace, the only exception we have is a very thin layered
> kernelmodule in order
> to improve the datatransfer.

Markus, can you point to some example so I can get a clearer picture?

Thanks,
   Antonio

-- 
Antonio Ospite
http://ao2.it

PGP public key ID: 0x4553B001

A: Because it messes up the order in which people normally read text.
   See http://en.wikipedia.org/wiki/Posting_style
Q: Why is top-posting such a bad thing?

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Thu, Nov 11, 2010 at 12:29 AM, Antonio Ospite
<ospite@studenti.unina.it> wrote:
> On Thu, 11 Nov 2010 00:13:09 +0100
> Markus Rechberger <mrechberger@gmail.com> wrote:
>
>> On Wed, Nov 10, 2010 at 11:48 PM, Mohamed Ikbel Boulabiar
>> <boulabiar@gmail.com> wrote:
>> > On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
>> > <ospite@studenti.unina.it> wrote:
>> >> If there are arguments against a kernel driver I can't see them yet.
> [...]
>> > If I want to use this device, I will add many userspace code to create
>> > the skeleton model and that need much computation. Kernel Module adds
>> > performance to my other code.
>>
>> just some experience from our side, we do have fully working
>> video4linux1/2 drivers
>> in userspace, the only exception we have is a very thin layered
>> kernelmodule in order
>> to improve the datatransfer.
>
> Markus, can you point to some example so I can get a clearer picture?
>

unfortunately we're closed source (and much more advanced) but you can
have a look at other projects:

* libv4l2
* freebsd has webcamd or something like that to emulate analog
tv/webcams and dvb (they are even reusing linux kernel drivers with a
userspace wrapper - so everything works in userspace for them).

aside of that you can just debug userspace drivers with gdb, valgrind
etc. if issues come up it will only affect your work not the entire
system, kernel is seriously something critical.

Markus
> Thanks,
>   Antonio
>
> --
> Antonio Ospite
> http://ao2.it
>
> PGP public key ID: 0x4553B001
>
> A: Because it messes up the order in which people normally read text.
>   See http://en.wikipedia.org/wiki/Posting_style
> Q: Why is top-posting such a bad thing?
>

Re: Bounty for the first Open Source driver for Kinect

[Mohamed Ikbel Boulabiar]

On Thu, Nov 11, 2010 at 12:36 AM, Markus Rechberger
<mrechberger@gmail.com> wrote:
> I've seen alot projects failing due not having enough users
> If it should mainly remain a hacker only project then a kernel module
> should be fine.
sorry ?

> aside of that you can just debug userspace drivers with gdb, valgrind
> etc. if issues come up it will only affect your work not the entire
> system, kernel is seriously something critical.
So you think that most of actual drivers which are inside the kernel are bad ?

if it is inside the kernel it will be better maintained and fixed.
External dependencies will break many things and add exceptions.
You already got an answer for an issue similar to this from Linus
Torvalds and Andrew Morton
http://lkml.org/lkml/2007/10/10/244

Most of people want to download a kernel that just has all things
inside. not search for other dependencies somewhere. If it is inside
the kernel, many licensing problems will disappear.
What if I want to develop something based on that userspace GPL
library, should I search which license should I use (and there are
many MIT/BSD/LGPL/...)?

i

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Thu, Nov 11, 2010 at 1:13 AM, Mohamed Ikbel Boulabiar
<boulabiar@gmail.com> wrote:
> On Thu, Nov 11, 2010 at 12:36 AM, Markus Rechberger
> <mrechberger@gmail.com> wrote:
>> I've seen alot projects failing due not having enough users
>> If it should mainly remain a hacker only project then a kernel module
>> should be fine.
> sorry ?
>
>> aside of that you can just debug userspace drivers with gdb, valgrind
>> etc. if issues come up it will only affect your work not the entire
>> system, kernel is seriously something critical.
> So you think that most of actual drivers which are inside the kernel are bad ?
>
> if it is inside the kernel it will be better maintained and fixed.
> External dependencies will break many things and add exceptions.
> You already got an answer for an issue similar to this from Linus
> Torvalds and Andrew Morton
> http://lkml.org/lkml/2007/10/10/244
>

As time goes on things evolve and change.

> Most of people want to download a kernel that just has all things
> inside. not search for other dependencies somewhere. If it is inside
> the kernel, many licensing problems will disappear.
> What if I want to develop something based on that userspace GPL
> library, should I search which license should I use (and there are
> many MIT/BSD/LGPL/...)?
>

I'm not writing about open vs closed source just that userspace is
more flexible for endusers (don't mix that up).

Markus
> i
>

Re: Bounty for the first Open Source driver for Kinect

[hermann pitton]

Am Donnerstag, den 11.11.2010, 00:36 +0100 schrieb Markus Rechberger:
> On Thu, Nov 11, 2010 at 12:29 AM, Antonio Ospite
> <ospite@studenti.unina.it> wrote:
> > On Thu, 11 Nov 2010 00:13:09 +0100
> > Markus Rechberger <mrechberger@gmail.com> wrote:
> >
> >> On Wed, Nov 10, 2010 at 11:48 PM, Mohamed Ikbel Boulabiar
> >> <boulabiar@gmail.com> wrote:
> >> > On Wed, Nov 10, 2010 at 10:24 PM, Antonio Ospite
> >> > <ospite@studenti.unina.it> wrote:
> >> >> If there are arguments against a kernel driver I can't see them yet.
> > [...]
> >> > If I want to use this device, I will add many userspace code to create
> >> > the skeleton model and that need much computation. Kernel Module adds
> >> > performance to my other code.
> >>
> >> just some experience from our side, we do have fully working
> >> video4linux1/2 drivers
> >> in userspace, the only exception we have is a very thin layered
> >> kernelmodule in order
> >> to improve the datatransfer.
> >
> > Markus, can you point to some example so I can get a clearer picture?
> >
> 
> unfortunately we're closed source (and much more advanced) but you can
> have a look at other projects:

Markus,

please go away with such.

Despite of all previously, this is _not_ a place for any closed source
to discuss. There is nothing to discuss on that, please stop it.

Either try to come back with open source in a new round, or at least
don't try to hide what you have. Without all the code and hardware
specific stuff _previously_ developed/hacked on v4l-dvb, you don't
exist.

I still admit, overall, you did a very good job previously, but all
others are _not_ just your captives after some clashes.

With "unfortunately we're closed source", you deliberately declare, that
you have nothing to do with open source at all anymore.

?

So, what is the remaining interest for you, except that you can continue
easier in userspace, instead of getting a hard block in the kernel, if
some enough have enough of your "closed source" ?

Cheers,
Hermann


> * libv4l2
> * freebsd has webcamd or something like that to emulate analog
> tv/webcams and dvb (they are even reusing linux kernel drivers with a
> userspace wrapper - so everything works in userspace for them).
> 
> aside of that you can just debug userspace drivers with gdb, valgrind
> etc. if issues come up it will only affect your work not the entire
> system, kernel is seriously something critical.
> 
> Markus
> > Thanks,
> >   Antonio
> >
> > --
> > Antonio Ospite
> > http://ao2.it

Re: Bounty for the first Open Source driver for Kinect

[Hans de Goede]

Hi,

On 11/10/2010 10:14 PM, Markus Rechberger wrote:
> On Wed, Nov 10, 2010 at 9:54 PM, Mohamed Ikbel Boulabiar
> <boulabiar@gmail.com>  wrote:
>> The bounty is already taken by that developer.
>>
>> But now, the Kinect thing is supported like a GPL userspace library.
>> Maybe still need more work to be rewritten as a kernel module.
>>
>
> This should better remain in userspace and interface libv4l/libv4l2 no
> need to make things more complicated than they have to be.
>

As the author and maintainer of libv4l I say no, webcam drivers and
the like belong in kernel space. libv4l is there to add things
like format conversion (de-bayering in this case) which do not belong
in userspace.

Also there is no way to do 100% reliable isoc data handling from
userspace.

Regards,

Hans

Re: Bounty for the first Open Source driver for Kinect

[Markus Rechberger]

On Thu, Nov 11, 2010 at 12:08 PM, Hans de Goede <hdegoede@redhat.com> wrote:
> Hi,
>
> On 11/10/2010 10:14 PM, Markus Rechberger wrote:
>>
>> On Wed, Nov 10, 2010 at 9:54 PM, Mohamed Ikbel Boulabiar
>> <boulabiar@gmail.com>  wrote:
>>>
>>> The bounty is already taken by that developer.
>>>
>>> But now, the Kinect thing is supported like a GPL userspace library.
>>> Maybe still need more work to be rewritten as a kernel module.
>>>
>>
>> This should better remain in userspace and interface libv4l/libv4l2 no
>> need to make things more complicated than they have to be.
>>
>
> As the author and maintainer of libv4l I say no, webcam drivers and
> the like belong in kernel space. libv4l is there to add things
> like format conversion (de-bayering in this case) which do not belong
> in userspace.
>
> Also there is no way to do 100% reliable isoc data handling from
> userspace.
>

That's just your opinion we have 100% reliable isoc data handling in
userspace, transferring 21 Mbyte/sec without any problem. And the
driver works from 2.6.15 on - without recompiling.
We're just about to release a new device in a few days, kernelsupport
is absolutely not interesting since most distributions would not ship
support for those devices at time of product release.

Best Regards,
Markus